Systematic insight into the active constituents and mechanism of Guiqi Baizhu for the treatment of gastric cancer

Abstract

Traditional Chinese medicine treatment of diseases has been recognized, but the material basis and mechanisms are not clear. In this study, target prediction of the antigastric cancer (GC) effect of Guiqi Baizhu (GQBZP) and the analysis of potential key compounds, key targets, and key pathways for the therapeutic effects against GC were carried out based on the method of network analysis and Kyoto Encyclopedia of Genes and Genomes enrichment. There were 33 proteins shared between GQBZP and GC, and 131 compounds of GQBZP had a high correlation with these proteins, indicating that the PI3K‐AKT signaling pathway might play a key role in GC. From these studies, we selected human epidermal growth factor receptor 2 (HER2) and programmed cell death 1‐ligand 1 (PD‐L1) for docking; the results showed that 385 and 189 compounds had high docking scores with HER2 and PD‐L1, respectively. Six compounds were selected for microscale thermophoresis (MST). Daidzein/quercetin and isorhamnetin/formononetin had the highest binding affinity for HER2 and PD‐L1, with K _d values of 3.7 μmol/L and 490, 667, and 355 nmol/L, respectively. Molecular dynamics simulation studies based on the docking complex structures as the initial conformation yielded the binding free energy between daidzein/quercetin with HER2 and isorhamnetin/formononetin with PD‐L1, calculated by molecular mechanics Poisson‐Boltzmann surface area, of −26.55, −14.18, −19.41, and −11.86 kcal/mol, respectively, and were consistent with the MST results. In vitro experiments showed that quercetin, daidzein, and isorhamnetin had potential antiproliferative effects in MKN‐45 cells. Enzyme activity assays showed that quercetin could inhibit the activity of HER2 with an IC₅₀ of 570.07 nmol/L. Our study provides a systematic investigation to explain the material basis and molecular mechanism of traditional Chinese medicine in treating diseases.

We are committed to establishing a systematic research method based on network pharmacology, multitarget molecular docking, molecular dynamics simulation, and protein and experimental verification in vitro and in vivo, to establish a systematic analysis method for traditional Chinese medicine (TCM) treatment of diseases. We aim to provide a possible theoretical and experimental basis for the standardization and internationalization of TCM.

Abbreviations

EGFR

epidermal growth factor receptor

GC

gastric cancer

GQBZP

Guiqi Baizhu prescription

HER2

human epidermal growth factor receptor 2

IL

interleukin

KEGG

Kyoto Encyclopedia of Genes and Genomes

LED

Light Emitting Diode

MD

molecular dynamics simulation

MMFFs

molecular mechanics force fields

MM‐PBSA

molecular mechanics Poisson‐Boltzmann surface area

MST

microscale thermophoresis

NF‐kB

nuclear factor ‐ kappa beta

PD‐1

programmed cell death 1

PDB

RCSB Protein Data Bank

PD‐L1

programmed cell death 1‐ligand 1

PPI

protein‐protein interaction

TCM

traditional Chinese medicine

TTD

Therapeutic Target Database

1. INTRODUCTION

Gastric cancer is one of the most common malignant tumors in the world. ¹ Due to the high heterogeneity of GC, it is a disease involving multiple genes and multiple targets, making treatment difficult. ² Therefore, it is urgent to explore new treatment methods. Blockers targeting PD‐L1, ³ PD‐1, ⁴ HER2, ⁵ and other targets have become an important direction in study of GC. Overexpression of HER2 in GC patients is often positively correlated with PD‐L1 expression, which could depend on the PI3K‐AKT‐mTOR pathway, ⁶ suggesting that the combination of drugs aimed at the above targets might be a new therapeutic pathway.

The multipoint fine adjustment, coordination, and efficacy enhancement effect of TCM have the characteristics of supporting the healthy factors and eliminating pathogenic factors, which should be advantageous in antitumor treatment. However, the current pharmacological evaluation method of TCM is still based on a single “disease gene‐target drug” research model. ⁷ The specific mechanism for effects of many TCM formulas on cancer still needs further investigation, given that the chemical composition of TCM is very complex. The further evaluation of the pharmacodynamic function of TCM compounds and to determine their mechanisms of action are the focus of further research.

Guiqi Baizhu prescription is modified from Huang Qi prescription in volume 272 of “Prescriptions for Universal Relief”. ⁸ Previous experiments showed that GQBZP could reduce liver ⁹ and kidney ¹⁰ functional damage induced by cisplatin, can improve the quality of life for patients with GC during chemotherapy, and regulate the immune balance. ¹¹ However, the molecular mechanism and material basis of its effectiveness remain unclear.

Molecular docking and network pharmacology are widely used to study the material basis and mechanism of TCM in the prevention and treatment of disease. ¹² , ¹³ , ¹⁴ Network pharmacology provides a new approach to systematically search for the active components and targets of TCM. Network pharmacology fits the multicomponent, multitarget dynamics of TCM and helps to identify the molecular mechanism of the overall effects of TCM. ¹⁵ Computer‐aided drug design, such as molecular docking and molecular dynamics simulations, is used to calculate the interaction and affinity between a compound and a target to generate the material basis and mechanism of action of TCM by virtual screening. ¹⁶ , ¹⁷ For protein‐ligand binding interactions, the MM‐PBSA method is usually applied to the calculation of binding free energies (ΔG_bind) of small molecule ligands bound to large biomolecule receptors. ¹⁸ Microscale thermophoresis is used to determine the affinity of the interacting biomolecules with high sensitivity. ¹⁹ Microscale thermophoresis can be used to identify appropriate compounds for cell experiments.

In this study, we identify key targets and potentially active compounds using network pharmacology, molecular docking, and MM‐PBSA calculation. We used cell research, MST, and enzyme activity analysis to verify the activity of selected compounds and thereby identify and characterize the material basis and mechanism of GQBZP in the treatment of GC.

2. MATERIAL AND METHODS

2.1. Identification of GQBZP components and targets, PPI analysis, and KEGG enrichment

2.1.1. Collection of compounds

The compounds of GQBZP were collected from Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP) ²⁰ databases and were supplemented through searches of published reports. The structures of compounds were confirmed in PubChem databases. ²¹

2.1.2. Target analysis of the anti‐GC effect of GQBZP

The simplified molecular input line entry specification (SMILES) code of each structure was obtained through PubChem. Swiss Target Prediction ²² was used to predict the potential targets of GQBZP. The drug targets for “GC” were retrieved by using “gastric cancer” and “gastric carcinoma” as keywords through TTD ²³ and Drug Bank. ²⁴ The common targets of GQBZP and GC were obtained by intersection.

2.1.3. Analysis of the key compounds of GQBZP against GC

Based on the identified anti‐GC targets of GQBZP, a herbs‐compounds‐targets network was constructed in Cytoscape 3.6.2, ²⁵ and the degree value of compounds in the network was used to analyze the principal effective compounds of herbs.

2.1.4. Kyoto Encyclopedia of Genes and Genomes enrichment and PPI analysis

The anti‐GC targets of GQBZP were imported into DAVID. ²⁶ The Select identifier was the official gene symbol, the List type was the gene list, the limited species was Homo sapiens, the threshold was P <.05, and KEGG pathway analysis was carried out. The online analysis platform Omicshare was used to visualize the enrichment path. ²⁷ String ²⁸ was used to analyze the PPI of the target, the species selected was Homo sapiens, and the comprehensive score for protein interaction was set as greater than 0.7 as the screening condition. Finally, the PPI information was imported into Cytoscape 3.6.2 for network analysis.

2.2. Molecular docking

2.2.1. Drug‐like properties analysis of GQBZP

Through collection of compounds in GQBZP, a compound database of molecular docking was established. Drug‐likeness was determined by Lipinski’s rule. ²⁹

2.2.2. Protein preparation

The 3D structure of HER2 (PDB ID: 3RCD) ³⁰ and PD‐L1(PDB ID: 4Z18) ³¹ were downloaded from PDB databases. ³² The binding pocket of HER2 was the substrate‐binding site and the active residues of HER2 were Lys753, Asp863, Thr862, Phe864, and Met801. ³⁰ The PD‐L1 binding pocket belonged to protein‐protein interface binding sites and the potential active residues were Tyr56, Met115, Ala121, Tyr123, and Asp122. ³¹ Finally, the Prep Wiz module was used to preprocess these targets in Schrödinger. The screen greedy box center read of HER2 was x cent −38.142190, y cent −29.102648, and z cent −74.036287, with the size of the box at 10 Å. The screen greedy box center read of PD‐L1 was x cent −6.006301, y cent −15.078195, and z cent −4.842327, with the size of the box at 10 Å.

2.2.3. Molecular docking

Compounds collected were prepared by the LigPrep module. ³² Molecular docking was carried out using the Glide module with the extra docking precision method (Glide XP) in Schrӧdinger 2020‐2. ³⁴ , ³⁵ The corresponding low‐energy conformation was obtained by the molecular mechanics force fields. Finally, Epik 28 ³⁶ used a pH value of 7.0 ± 2.0 to conditionally allocate the ionization state and undertake docking calculations. The interaction mechanism between drug components and protein was studied by Pymol. ³⁷

2.3. Microscale thermophoresis assay

The TCM compounds’ affinity for proteins was measured by MST Reagent (NanoTemper Technologies) and Monolith NT.115 (NanoTemper Technologies). The proteins (Human PD‐L1 Fc Protein, PeproTech; and Recombinant Human Her2/ErbB2 Protein, Sino Biological) were desalted into MST buffer (10 mmol/L HEPES, pH 7.5, 150 mmol/L NaCl) before the experiment. The proteins were fluorescently labelled and the protein concentration was adjusted to 10 μmol/L. Fluorescent dye NT‐647‐NHS was added, mixed, and incubated for 30 min at 25°C in the dark. For each assay, the labelled protein (approximately 0.1 μmol/L) was mixed with the same volume of unlabeled ligands quercetin (Sigma‐Aldrich), daidzein (Sigma‐Aldrich), isorhamnetin (Sigma‐Aldrich), or formononetin (Saitong) in 16 serial concentrations at room temperature. The samples were then loaded into premium capillaries and measured at 25°C by using 20% Light Emitting Diode power and medium MST power. Data analyses were undertaken using MO Affinity Analysis 2.2.4 software. With a confidence of 68%, the K _d value is within the given range.

2.4. Molecular dynamics

2.4.1. System setup

The initial structure of the proteins complex with compounds was obtained from molecular docking. Structure optimization and frequency calculations for the small molecule quercetin, daidzein, isorhamnetin, and formononetin were determined at the HF/6‐31G* level of Gaussian 09. The restrained electrostatic potential was used to describe the partial atomic charges. The parameters for the small molecules were generated from Amber force field. ³⁸ The standard Amber force field for bioorganic systems ³⁹ was used to describe the protein. The counter ions were added to neutralize each system. All systems were solvated using an atomistic TIP3P ³⁹ water box with at least 10 Å distance around the complex.

2.4.2. Molecular dynamics simulations

Amber20 ⁴⁰ was used throughout the simulation process and the force field FF14SB was selected to describe the protein. The energy minimization was first carried out with the steepest descent method and switched to conjugate gradient every 2500 steps for 5000 steps with a 0.1 kcal/mol·Å² restraint on all atoms of the complexes. Following this step, another two rounds of energy minimization were undertaken by restraining only protein and further releasing all the restraints for 5000 steps of each round. Long‐range Coulombic interactions were handled using the particle mesh Ewald (PME) summation. ³⁹ For the equilibration and subsequent production runs, the SHAKE algorithm ⁴¹ was used on all atoms covalently bonded to a hydrogen atom, allowing for an integration time step of 2 fs. The systems were gently annealed from 0 to 310 K over a period of 50 ps using a Langevin thermostat with a coupling coefficient of 1.0/ps and a force constant 2.0 kcal/mol·Å² on the complex. All subsequent stages were carried out in the isothermal isobaric (NPT) ensemble using a Berendsen barostat with a target pressure of 1 bar. An additional five rounds of MD (100 ps each at 310 K) were carried out with decreasing restraint weights from 2.0, 1.5, 1.0, 0.5 to 0.1 kcal/mol·Å². By releasing all the restrains, the system was again equilibrated for 500 ps. The production phase of the simulations was run without any restrains for a total of 50 ns.

2.4.3. Binding free energy calculations

The MM‐PBSA calculation was carried out using Amber20. ⁴² The first step of the MM‐PBSA method is the generation of multiple snapshots from a molecular dynamics trajectory of the protein‐ligand complex, which was stripped of water molecules and counter ions. Two hundred and fifty snapshots, equally spaced at 20 ps intervals, were extracted from the equilibration section (last 10 ns) of the molecular dynamics trajectory. For each snapshot, the free energy was calculated for each molecular species (complex, protein, and ligand). The binding free energy was computed as the difference ⁴² :

$$\mathrm{\Delta }{G}_{\text{bind}}=G_{\text{complex}}‐G_{\text{protein}}‐G_{\text{ligand}}$$

2.5. Enzyme activity assays of HER2

All enzymatic reactions were carried out at 30°C for 40 min. The 50 µl reaction mixture contained 40 mmol/L Tris, 10 mmol/L MgCl₂, 0.1 mg/ml BSA, 1 mmol/L DTT, 10 µmol/L ATP, 0.4 ng/µl HER2 kinase, and 100 µmol/L lipid substrate. The compounds were diluted in 10% DMSO and 5 µl of the dilution was added to a 50 µl reaction; the final concentration of DMSO was 1% in all reactions. These assays were undertaken using the HTRF kinEASE TK kit (Cisbio). The fluorescent signals were correlated with the amount of ATP present and were inversely correlated with kinase activity. The IC₅₀ values were calculated using nonlinear regression with normalized dose‐response fit.

2.6. Screening, culture, and cytotoxicity detection of GC cells

Human GC MKN‐45, SGC‐7901, and BGC‐823 cell lines and the human gastric mucosal epithelial GES‐1 cell line were cultured in DMEM (Hyclone) with 10% FBS, 1% penicillin, and 1% streptomycin, incubated in 5% CO₂.

Protein (50 μg) from the MKN‐45, SGC‐7901, BGC‐823, and GES‐1 cell lines was harvested from RIPA lysis buffer and separated by 10% SDS‐PAGE. Protein was transferred to a PVDF membrane (Millipore). After blocking with 10% nonfat milk, membranes were incubated with primary Abs for PD‐L1 (GeneTex) or HER2 (Abcam). Membranes were incubated with secondary Abs followed by enhanced chemiluminescence reagents (Yeasen). The GAPDH protein was used as a reference. ⁴³ The GC cells were incubated with HER2 Ab (1:250) or PD‐L1 (1:200) diluted in PBS containing 1% BSA and incubated overnight at 4°C, followed by addition of fluorescent‐labeled secondary Abs (Alexa Fluor 647; Abcam) for 1 h at room temperature. ⁴⁴

MKN‐45 cells with approximately 6 × 10³ cells/well were plated in a 96‐well standard microplate in DMEM with 10% FBS, 1% penicillin, and 1% streptomycin, and incubated in 5% CO₂. After the cells became adherent, the medium was replaced and different doses of compounds were added. Cells were exposed to various concentrations (five serial dilutions in triplicate) of quercetin, daidzein, isorhamnetin, and formononetin. After 72 h of culture, the medium was removed and 100 µl DMEM, supplemented with 10% CCK‐8, was added to each well and incubated in the dark for an additional 2 h at 37°C. The absorbance was then measured using microplate reader (iMark; Bio‐Rad) at a wavelength of 450 nm. The cell viability was expressed as a percentage of the control values (blank). The experiments were done in triplicate.

2.7. Statistical analysis

Differences among three or more experimental samples were analyzed by ANOVA. Significance was set at P <.05.

3. RESULTS

3.1. Identification of GQBZP components and targets, PPI analysis, and KEGG enrichment

The seven herbs in GQBZP all belong to the spleen meridian, which has the characteristics of both attacking and supplementing by strengthening the spleen and stimulating blood flow. Information on each herb is shown in Table 1. In this study, 787 compounds were collected, which were mainly flavonoids, alkaloids, and coumarins. Their structures are shown in Figures S1‐S61.

TABLE 1.

Composition and characteristics of each constituent of Guiqi Baizhu prescription

Drug name/English name	Medicinal part	Channel tropism	Medicinal efficacies	Role in the prescription
Astragali radix	Root, rhizome	Lung, spleen	Invigorate the Zhong qi, detoxification, purulence, diuresis	Monarch
Atractylodis macrocephalae	Root, rhizome	Spleen, lung, spleen	Tonify spleen and replenish qi	Monarch
Angelicae	Rhizome	Heart, spleen, liver	Replenish blood and activate blood	Minister
Paeoniae radix alba	Root	Spleen, liver	Nourish blood and relieve pain	Minister
Pericarpium citri reticulatae	Mature pericarp	Spleen, lung	Regulate qi, invigorate spleen, and dry dampness	Assistant
Rhubarb	Root, rhizome	Spleen, stomach, large intestine, liver, pericardium	Promote the excretion of toxins and evil, regulate qi and stagnation, invigorate the spleen and stomach, and activate qi to promote blood circulation	Assistant
Glycyrrhizae	Root, Rhizome	Heart, spleen, lung, spleen	Invigorate spleen and replenish qi, pain relief, and harmonize various medicines	Messenger

The potential targets of the seven herbs of GQBZP were predicted by the Swiss Target Prediction database, and 885 targets were obtained after removing duplicates. One hundred and twenty nine GC potential targets were obtained from the Drug Bank and TTD database after removing duplicates, and 33 key potential targets were finally obtained after intersection. The herb‐compound‐target network consists of 171 nodes and 938 edges (Figure 1A), with green, blue, and purple representing herbs, compounds, and key targets, respectively. The top 10 compounds are kaempferol, naringenin, jaranol, mairin, quercetin, isorhamnetin, (3S,8S,9S,10R,13R,14S,17R)‐10,13‐dimethyl‐17‐[(2R,5S)‐5‐propan‐2‐yloctan‐2‐yl]‐2,3,4,7,8,9,11,12,14,15,16,17‐dodecahydro‐1H‐cyclopenta[a]phenanthren‐3‐ol, sitosterol, beta‐sitosterol, and 1,3‐dihydroxy‐8,9‐dimethoxy‐6‐benzofurano[3,2‐c] chromenone. HSP90AA1 (66), EGFR (66), KDR (62), PTGS1 (62), SRC (61), CA9 (52), MET (48), TOP1 (37), ODC1 (34), and AKT1 (29) are the potential key targets of GQBZP (Table 2). The PPI network (Figure 1B) was constructed with highly reliable (more than 0.7) targets of interaction, which includes 22 nodes and 41 edges; in the network, circles represent targets, and the darker the color and bigger the size, the greater the degree value. In the protein interaction network of GQBZP, VEGFA (11), HSP90AA1 (10), SRC (9), EGFR (8), and FGF2 (6) are the top five nodes, which could play important roles in the treatment process.

FIGURE 1.

Identification of Guiqi Baizhu prescription (GQBZP) components and targets, protein‐protein interaction (PPI) analysis, and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment. A, Traditional Chinese medicine‐compound‐target network of GQBZP. B, PPI network of targets of GQBZP. C, KEGG analysis of potential targets. The bubble size represents the number of genes enriched in this pathway, the bubble color difference represents the level of gene enrichment in this pathway, and the circle size represents the number of targets contained in the pathway. The redder the color, the smaller the P value. VEGF, vascular endothelial growth factor

TABLE 2.

Key targets of Guiqi Baizhu prescription in the treatment of gastric cancer

No.	Gene name	Uniprot ID	Protein name
1	HSP90AA1	P07900	Heat shock protein 90 alpha family class A member 1
2	EGFR	P00533	Epidermal growth factor receptor
3	KDR	P35968	Vascular endothelial growth factor receptor 2
4	PTGS1	P23219	Cyclooxygenase‐1
5	SRC	P12931	Tyrosine‐protein kinase SRC
6	CA9	Q16790	Carbonic anhydrase IX
7	MET	P08581	Hepatocyte growth factor receptor
8	TOP1	P11387	DNA topoisomerase I
9	ODC1	P11926	Ornithine decarboxylase
10	AKT1	P31749	Serine/threonine‐protein kinase AKT

The potential protein targets could play an important role in GC, but whether they are a drug target and can be targeted for drug development still requires practical detection. Among the HER family, HER2 appears as a keystone for ERBB signaling and is a key feature for cross‐talk between numerous cell signaling pathways. ⁴⁵ Both EGFR and HER2 are closely related to the abnormal proliferation of tumor cells, and the abnormal expression of HER2 in tumor tissues promotes the proliferation of tumor cells. ⁴⁶ At present, there are many studies on EGFR/HER2 dimers, and trastuzumab targeting HER2 has been approved for clinical treatment of GC. Therefore, HER2 was selected as the key target of molecular docking.

The enrichment of the KEGG pathway of GQBZP in the treatment of GC is shown in Figure 1C. The result showed that the PI3K‐AKT signaling pathway might play a role in the treatment of GC using GQBZP. Overexpression of HER2 in GC patients is often positively correlated with PD‐L1 expression, which could depend on the PI3K‐AKT‐mTOR pathway, ⁶ suggesting that the combination of drugs aimed at the above targets could be a promising treatment strategy. Suh et al ⁴⁷ reported that PD‐L1 expression is positively correlated with EGFR/HER2 signaling pathway activation in seven GC cell lines. Targeted multichannel combination therapy can further improve the immune response and survival rate, including the combination of targeted therapy, chemotherapy, and antiangiogenic therapy. ⁴⁸ Therefore, we believe that PD‐L1 overexpression is related to the EGFR/HER2 pathway, and drugs targeting these pathways could reduce the inhibition of antitumor immunity. According to a report released by the Cancer Research Institute, ⁴⁹ there were 1176 combined therapy studies in 2018, of which 21 were PD‐1/PD‐L1 inhibitors combined with HER2. Therefore, based on the results of network pharmacology, our previous study, and other published reports, HER2 and PD‐L1 were selected as therapeutic targets to explore the effect and possible molecular mechanisms of multitarget combination therapy for GC.

3.2. Molecular docking

In this study, 787 compounds were collected, 483 of which conformed to Lipinski’s rule (Table 3). The structure of HER2 and PD‐L1 proteins are shown in Figure 2.

TABLE 3.

Number of Guiqi Baizhu prescription compounds before and after drug‐like identification and after docking with human epidermal growth factor receptor 2 (HER2)/programmed cell death 1‐ligand 1 (PD‐L1) protein (≤−4)

	Compounds	Drug‐like properties	Docking with HER2	Docking with PD‐L1
Astragali radix	87	37	33	19
Atractylodis macrocephalae	55	36	29	10
Angelicae	125	89	69	25
Paeoniae radix alba	85	40	28	8
Pericarpium citri reticulatae	63	51	26	5
Rhubarb	92	44	37	17
Glycyrrhizae	280	186	163	95
Total	787	483	385	189

FIGURE 2.

Crystal structure and amino acid sites of target proteins. A, Structure of human epidermal growth factor receptor 2 (HER2). B, Amino acid sites of HER2. C, Structure of programmed cell death 1‐ligand 1 (PD‐L1). D, Amino acid sites of PD‐L1

After docking, scores were obtained and expressed in absolute values, and the small molecules with high scores in different herbs are shown in Table 3. The results show that 385 compounds having drug‐like properties in GQBZP had a docking score of −4 or lower with HER2, and 189 had a docking score of −4 or lower with PD‐L1. Table 4 and Table 5 list the small molecules with higher scores after docking with the target proteins. The docking results show that Astragali radix, Angelicae, and rhubarb in GQBZP contained potential active ingredients, which corresponds with the effects of Astragali radix in prescriptions on invigorating qi and the spleen, Angelicae in nourishing blood and replenishing blood, and rhubarb in promoting blood circulation. From the two‐way effect of regulating immunity and treating tumors, it preliminarily explains that GQBZP might play a coordinated role in the treatment of GC through multiple targets, such as HER2 and PD‐L1. ⁵⁰

TABLE 4.

Compounds with higher scores for docking with human epidermal growth factor receptor 2 in Guiqi Baizhu prescription

TCMSP ID	TCM drug	Compound name	Docking score	MW	HBA	HBD	logpo/W	Rotor
MOL000030	Atractylodis macrocephalae	Acridine	−9.743	148.204	1.70	1	2.870	3
MOL000038	Atractylodis macrocephalae	Akridin	−9.181	179.221	1.00	0	3.403	0
MOL000040	Atractylodis macrocephalae/Angelicae/Glycyrrhizae	Scopoletol	−8.488	192.171	4.00	1	1.021	2
MOL000044	Atractylodis macrocephalae	Atractylenolide II	−7.960	232.322	3.00	0	2.666	0
MOL000098	Astragali radix/Glycyrrhizae	Quercetin	−7.788	302.240	5.25	4	0.472	5
MOL000106	Paeoniae radix alba	Pyrogallol	−7.780	126.112	2.25	3	0.103	3
MOL000354	Astragali radix/Glycyrrhizae	Isorhamnetin	−7.533	316.267	5.25	3	1.402	5
MOL000259	Pericarpium citri reticulatae/Angelicae	Carvacrol	−7.583	150.220	0.75	1	3.291	2
MOL000390	Astragali radix	Daidzein	−7.331	254.242	4.00	2	1.845	3
MOL000396	Astragali radix	(+)− Syringaresinol	−7.275	418.443	7.90	2	3.387	6

HBA, hydrogen bond acceptors; HBD, hydrogen bond donors; MW, molecular weight; logPo/w, lipid‐water partition coefficient; Rotor, number of rotatable bonds; TCM, traditional Chinese medicine; TCMSP, Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform.

TABLE 5.

Compounds with higher scores for docking with programmed cell death 1‐ligand 1 in GQBZP

TCMSP ID	TCM drug	Compound name	Docking score	MW	HBA	HBD	logPo/w	Rotor
MOL000030	Atractylodis macrocephalae	Acridine	−5.575	148.204	1.70	1	2.870	3
MOL000040	Atractylodis macrocephalae/Angelicae/Glycyrrhizae	Scopoletol	−5.417	192.171	4.00	1	1.021	2
MOL000098	Astragali radix/Glycyrrhizae	Quercetin	−5.215	302.240	5.25	4	0.472	5
MOL000106	Paeoniae radix alba	Pyrogallol	−5.217	126.112	2.25	3	0.103	3
MOL000114	Astragali radix	Vanillic acid	−5.198	168.149	3.50	2	1.042	3
MOL000354	Astragali radix/Glycyrrhizae	Isorhamnetin	−4.998	316.267	5.25	3	1.402	5
MOL000389	Astragali radix/Atractylodis macrocephalae	Ferulic acid (CIS)	−4.96	194.187	3.5	2	1.586	5
MOL000392	Astragali radix/ Glycyrrhizae	Formononetin	−4.883	268.268	4.0	1	2.698	3
MOL000414	Astragali radix	Caffeic acid	−4.876	180.160	3.5	3	0.636	5
MOL000416	Astragali radix	Lariciresinol	−4.865	360.406	6.4	3	2.916	8

HBA, hydrogen bond acceptors; HBD, hydrogen bond donors; MW,molecular weight; logPo/w,lipid‐water partition coefficient; Rotor,number of rotatable bonds; TCM, traditional Chinese medicine; TCMSP, Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform.

3.3. Microscale thermophoresis analysis and molecular dynamics

Six compounds that targeted HER2 and PD‐L1 were selected for MST experiments after docking. More than 12 concentration gradients were used to fit the K _d value of the equilibrium dissociation constant (Figure 3). Daidzein/quercetin and isorhamnetin/formononetin had the highest binding affinity for HER2 and PD‐L1, with K _d values of 3.7 μmol/L and 490, 667, 355 nmol/L, respectively, indicating a strong potential targeting ability.

FIGURE 3.

Results of microscale thermophoresis. A, Quercetin vs human epidermal growth factor receptor 2 (HER2). B, Daidzein vs HER2. C, Formononetin vs programmed cell death 1‐ligand 1 (PD‐L1). D, Isorhamnetin vs PD‐L1

To further study the interaction pattern of the above four compounds at the molecular level, MD of the corresponding complexes was carried out. As shown in Figure 4, for HER2, during a 50 ns MD the interaction modes were stable, as indicated by comparing the initial and final structures of the protein‐ligand complex, and the interaction between binding pocket residues was relatively conservative. ⁵¹ , ⁵² The interactive residues were Met801, Lys753, and Asp863, and hydrogen bonding was the main interaction (Figure 4A‐D). In the process of binding, the interactive amino acids of the binding pocket with small molecules required conformational adjustment of the molecules rather than the protein. For example, quercetin (Figure 4A,B) and daidzein (Figure 4C,D) added new hydrogen bonding with Gln799 and Cys805, respectively.

FIGURE 4.

Interaction mode of compounds with the target proteins. A, Quercetin vs human epidermal growth factor receptor 2 (HER2) before molecular dynamics simulation (MD). B, Quercetin vs HER2 after MD. C, Daidzein vs HER2 before MD. D, Daidzein vs HER2 after MD. E, Formononetin vs programmed cell death 1‐ligand 1 (PD‐L1) before MD. F, Formononetin vs PD‐L1 after MD. G, Isorhamnetin vs PD‐L1 before MD. H, Isorhamnetin vs PD‐L1 after MD

For PD‐L1, during a 50 ns MD, the interaction mode of isorhamnetin and PD‐L1 was stable and isorhamnetin interacted with Glu58 through hydrogen bonding (Figure 4E,F). The binding conformation of formononetin with PD‐L1 changed after MD (Figure 4G,H). The binding pocket of PD‐L1 is located in a plane composed of β‐sheets, which means that the structure is inflexible and the amino acids in the binding pocket cannot adjust in the binding of small molecules. Thus, a conformational adjustment of small molecules to better adapt to the pocket is necessary to achieve better binding. The chromone moiety of formononetin is closer to Ala121, Asp122, and Tyr123 after a 50 ns MD (Figure 4H). Both MM‐PBSA and MST showed that formononetin had higher binding free energy and affinity with PD‐L1, indicating that formononetin can achieve a better binding pattern by conformational adjustment. The pocket is made up of Glu58, Arg113, Ala121, Asp122, and Tyr123, which are key residues blocking PD‐1/PD‐L1 interaction. ³¹ , ⁵³ Zak et al ⁵⁴ reported that BMS‐202 and BMS‐8, as PD‐1/PD‐L1 blockers, could interact with PD‐L1 Ala121, Asp122, and Tyr123.

By conformational analysis, in summary, quercetin and formononetin are two key compounds in GQBZP and have strong ability to target HER2 and PD‐L1 with ideal modes of action.

Binding free energies between compounds and proteins were calculated by the MM‐PBSA method after MST, and the results are shown in Table 6. The binding free energies of quercetin and daidzein with HER2 are −26.55 kcal/mol and −14.18 kcal/mol, respectively, and the binding free energies of formononetin and isorhamnetin with PD‐L1 are −19.41 kcal/mol and −11.86 kcal/mol, respectively, which is consistent with the MST assays.

TABLE 6.

Binding free energies and energy components calculated by the molecular mechanics Poisson‐Boltzmann surface area method (kcal/mol)

	Quercetin	Daidzein	Formononetin	Isorhamnetin
HER2	−26.55	−14.18	no data	no data
PD‐L1	no data	no data	−19.41	−11.86

HER2, human epidermal growth factor receptor 2; PD‐L1, programmed cell death 1‐ligand 1.

3.4. Compounds targeting HER2 and PD‐L1 suppressed cell proliferation

Although we have confirmed that the compounds have high affinity with the target proteins, whether they have inhibitory effects still needs verification. We measured the effect of these compounds on the proliferation of GC cells. First, we used western blot to detect the expression of HER2 and PD‐L1 in the MKN‐45, SGC‐7901, and BGC‐823 GC cell lines and gastric mucosal epithelial GES‐1 cell line. The results showed that HER2 and PD‐L1 are highly expressed in MKN‐45 cells (Figure 5A,B). We used confocal microscopy to verify this result, which is consistent with western blot analysis (Figure 5C,D).

FIGURE 5.

Screening results of cell lines expressing human epidermal growth factor receptor 2 (HER2) and programmed cell death 1‐ligand 1 (PD‐L1). A, Western blot analysis of HER2 and PD‐L1 in MKN‐45, SGC‐7901, BGC‐823, and GES‐1 cells. B, Relative HER2 and PD‐L1 expression levels were quantified by normalization to GAPDH. ns, not significant. ****P <.0001. C, D, Images of HER2 and PD‐L1 immunofluorescence in MKN‐45 cells. Scale bar = 50 μm

As we had confirmed that HER2 and PD‐L1 were overexpressed in MKN‐45 cells, the cytotoxic activities of compounds targeting HER2 and PD‐L1 ⁵⁵ were determined by CCK‐8 assay; the IC₅₀ values of quercetin, daidzein, and isorhamnetin were 19.27 μmol/L, 21.48 μmol/L, and 10.84 μmol/L, respectively. Formononetin could not inhibit proliferation of tumor cells, as shown in Figure 6. Studies have shown that formononetin can block the activation and translocation of nuclear factor ‐ kappa beta (NF‐kB) in INS‐1 cells induced by IL‐1β. It has a good antiinflammatory effect, which could also include immunoregulation through this pathway. ⁵⁶ Formononetin and some of its metabolites could play an immunoregulatory role by enhancing the production of IL‐4 in T cells. ⁵⁷ Wang et al ⁵⁸ found that isorhamnetin can enhance innate immunity by interfering in an immune‐damaged mouse model. Shi et al ⁵⁹ also found that isorhamnetin can prevent or treat chronic inflammation, autoimmune diseases, and transplant rejection by regulating the role of cytokines in dendritic cells. ⁶⁰ This group found that isorhamnetin and formononetin, the active ingredients in Astragali radix, have immunoregulatory effects, through the study of network pharmacology. Our cell experiments yielded different results, however, which warrants further exploration.

FIGURE 6.

Inhibitory effects of compounds on MKN‐45 cell. A, Quercetin. B, Daidzein. C, Formononetin, D, Isorhamnetin

3.5. Quercetin suppressed HER2 kinase

According to the molecular docking analysis and cell proliferation assays, target prediction suggested that quercetin might be a potential HER2 kinase inhibitor. To verify this and explore the mechanisms of the compounds, the inhibitory effects of quercetin against the HER2 kinase were determined. The results show that quercetin strongly suppresses HER2 kinases, with an IC₅₀ value of 570.07 nmol/L, as shown in Figure 7.

FIGURE 7.

Enzyme inhibitory activity of human epidermal growth factor receptor 2

Quercetin impedes the propagation of various types of cancers, such as lung, prostate, liver, breast, colon, and cervical cancer. ⁶¹ Quercetin functions through various mechanisms involved in cell signaling and has the ability to inhibit carcinogen‐activating enzymes, showing anticancer effects based on binding to cell receptors and proteins. ⁶² , ⁶³ Furthermore, quercetin has been recently reported to have synergistic effects when combined with chemotherapeutic agents such as cisplatin, which could further improve the outcomes of traditional chemotherapy. ⁶⁴ Although compared with a positive drug, the docking scores and affinity of compounds of TCM with target protein (HER2) might be lower, the probability of drug resistance to those compounds might also be lower, and compared with mAbs, these compounds have the advantages of short half‐life, minimal side‐effects, and low cost. ⁶⁵ In addition, quercetin can cooperate with multiple compounds to avoid the adverse effects of strong binding of a single molecule. At the same time, quercetin can be used as a skeleton to modify its structure, or on this basis to screen other compounds of TCM based on structure matching. Moreover, the results of quercetin derivatives reported from 2012 to 2016 revealed that quercetin derivatives had higher biological activity, including higher cytotoxicity to breast cancer cells, HeLa cells, HepG2 cells, and other cells, than quercetin. ⁶⁶

The therapeutic effect of compounds of TCM might not be better than that of mAbs, but they have the advantages of high oral bioavailability and few side‐effects, and so have great prospects for clinical development. Quercetin and formononetin in GQBZP have stronger affinity for HER2 and PD‐L1, respectively, and could be the key compounds in the prescriptions for two‐way treatment of cancer by inhibiting the proliferation of cancer cells and improving the immune function of the body.

4. DISCUSSION

Traditional Chinese medicine is a comprehensive medicinal system, which has potential utilization based on thousands of years in clinical practice in China, providing the characteristics of multicomponents and multitargets and presenting a potent synergistic effect for treatment of various diseases with fewer side‐effects. ⁶⁷ The effectiveness of TCM in treating diseases has been recognized worldwide, ⁶⁸ , ⁶⁹ , ⁷⁰ but the material basis, potential mechanisms, interaction relationships, and modes of action of TCM and possible targets are still unclear, which limits the precise use and modern development of TCM. The above results show that, although our results do not have higher docking scores and binding affinities than those of approved mAbs or small molecule inhibitors, we preliminarily verify the theory of “synergistic pharmacological effect with multicomponents and multitargets” of TCM using network pharmacology, multitarget molecular docking, molecular dynamics simulation, and protein and cell experiments in vitro, and explain the possible material basis of GQBZP in the treatment of GC. Moreover, the clinical use of TCM can benefit from the synergistic effects of the ingredients, reduce adverse reactions, and effects of single drug treatment, and has better clinical application value. However, this study is limited to in vitro experiments and it is still necessary to further explore the potential activity of the active compounds of GQBZP with experimental studies in vivo to reveal the molecular mechanisms of its clinical therapeutic effect on GC. We are committed to establishing a systematic research method based on network pharmacology, multitarget molecular docking, molecular dynamics simulation, and protein and experimental verification in vitro and in vivo, to establish a systematic analysis method for TCM treatment of diseases, so as to provide a possible theoretical and experimental basis for the standardization and internationalization of TCM.

CONFLICT OF INTEREST

The authors have no conflict of interest.

Supporting information

Supplementary Material

Li L, Jin X‐J, Li J‐W, et al. Systematic insight into the active constituents and mechanism of Guiqi Baizhu for the treatment of gastric cancer. Cancer Sci. 2021;112:1772–1784. 10.1111/cas.14851